System virtualization is a technology which can divide a single host (e.g., computer, server, etc.), into multiple parts, or partitions, each running a separate instance, or image, of an operating system. The instances of the operating systems or partitions are separate, or isolated, from each other in some ways. For example, the partitions have separate file systems, separate users, separate applications, and separate processes. However, the partitions may also share some resources of the host. For example, the partitions can share the memory, the kernel, the processors, the hard drives, and/or other software, firmware, and/or hardware of the host. Thus, each partition or instance of the operating system can look and feel like a separate server or machine from the perspective of its users. These instances are commonly referred to as “virtual” or “virtualized” machine, and each partition may be referred to as a logical partition (LPAR).
One server or data processing system can generally host a number of LPARs. These LPARs generally have virtual resources assigned to them (e.g., virtual processors) which provide an abstraction of the physical resource from which it is based (e.g., the physical processor). For example, a virtual resource may be considered a logical entity that is backed up by the physical entity. In a dedicated mode of LPAR operation, physical resources are assigned as a whole to an LPAR such that the assigned physical resources are not shared by other LPARs. In a shared mode of LPAR operation, the physical resources are part of a pool of resources which are shared among the LPARs. Additionally, LPARs can be configured to have a certain entitlement capacity representing a guaranteed physical resource capacity for the LPAR. LPARs may also be configured as capped or uncapped. In a capped mode, the resource capacity for the LPAR is capped at some value (e.g., its entitlement capacity). In an uncapped mode, the LPAR can exceed its entitlement capacity when other shared resources are available.